Earthmover ERA5, Icechunk + Zarr
u100 and v100 (the 100m wind components) are read straight off a public, anonymous S3 bucket (versioned Zarr arrays via Icechunk), while the pipeline pulls only the India window it needs.
us-east-1 · CC-BY 4.0India is betting its energy future on wind. I compared two periods in the record and saw a small but clear change:
The data covers two decades of estimated wind speeds across India, measured about 100 metres above the ground: roughly the height of a modern wind turbine. The story moves from the national trend to the places where useful wind remains.
The yearly averages show a small decline: India was calmer in 2015–2025 than in 2005–2014. Looking at yearly averages helps separate that longer change from the much larger swing between the monsoon and the rest of the year. The chart lets you inspect each year and see how steady, or uneven, the change really was.
The difference is unlikely to be just a quirk of a few years. A Welch two-sample test on the annual averages gives p ≈ 0.03, which is below the usual 5% threshold. In other words, the later period is significantly calmer on average. That result describes India's land average; it does not mean that every place became calmer, or that the decline will continue at the same rate.
The national average hides the geography. Some places became windier, while many others became calmer. The map shows that difference for each land cell, so the broad decline does not get mistaken for an equal change everywhere.
Even with the decline, the strongest winds are still found in a few familiar places. They gather along parts of the coast and in mountain gaps. The map opens near Muppandal in the Kanyakumari corridor, where the pattern is easy to see. The tool below shows which land cells still pass a useful wind-speed threshold.
Wind power rises quickly as wind speed rises. A place with wind that is 20% faster can produce roughly 70% more energy per square metre of rotor. Raise the threshold and the best locations quickly gather back into the same few corridors.
The biggest changes still happen with the seasons. Wind builds before the southwest monsoon, peaks in the summer, and eases through winter. That seasonal swing is much larger than the long-term decline, which is why the slower change is easy to miss.
The direction of the wind adds another clue. In the windiest month, the fastest winds mostly come from a small set of directions. The wind rose gives a quick view of that pattern.
The analysis uses the public Earthmover ERA5 archive, which estimates wind conditions around the world. For India, the data was sampled once a day at about 100 metres above the ground, averaged across land, and then grouped by month and year. The technical details are below for anyone who wants to follow the process from the original data to the maps and charts.
u100 and v100 (the 100m wind components) are read straight off a public, anonymous S3 bucket (versioned Zarr arrays via Icechunk), while the pipeline pulls only the India window it needs.
us-east-1 · CC-BY 4.0xarray and zarr slice out the India bounding box (lat 6.5–37.5, lon 68.0–97.5) directly from the array: one hour for the map and twenty years for the time series.
125×119 grid · 20yr hourlyspeed = √(u² + v²), spatially averaged and resampled to monthly means; grouped by calendar month for the climatology envelope. The climatological peak month picks which day's wind field the map renders.
numpy · pandas · uvSummary JSON/CSV files and chunked arrays land in the static folder; the dashboard reads them at load time and renders the maps and charts entirely client-side.
static files · browser-rendered